Mezmo Open-Sources AURA, an SRE Agent Platform for Production Incident Investigation

Mezmo has released AURA, a production-tested Site Reliability Engineering (SRE) agent platform designed to automate incident investigation and resolution. The tool is now available under the Apache License, Version 2.0, allowing developers to deploy it in minutes on Linux or macOS via a simple install script.

AURA functions as a middleware layer that connects large language models to production observability tools. It handles the complex infrastructure required for agentic workflows, including guardrails, state management, streaming, and failure handling. By integrating with existing stacks such as Mezmo traces, Prometheus, and Kubernetes, AURA enables a team of specialist agents to correlate data points during incidents.

In a demonstrated workflow, AURA investigated a payment failure by analyzing latency metrics and deployment history. The agents identified an N+1 regression introduced by a specific service version, recommended a rollback to a previous stable version, and provided recovery checks for engineering follow-up. This approach allows SRE teams to move from manual log correlation to automated, evidence-based recommendations.

The platform is not a theoretical prototype; it began as Mezmo’s internal harness for operating its own SaaS platform and currently powers thousands of agent sessions monthly. For developers, AURA offers a guided setup process to connect existing infrastructure, after which agents can be customized or new ones added to fit specific SRE workflows. The project also emphasizes security, providing documentation on telemetry defaults, permission boundaries, and supply-chain verification for the downloaded artifacts.